TOP BYTE: Ahead of Biden's AI Order, AI Companies Announce New Security Coalition

On Monday, U.S President Joe Biden is expected to issue an executive order that will, among other things, require AI models to undergo assessments before being able to be used by federal workers. Other aspects of the announcement have yet to be confirmed, but reports state the focus of the order will be: "Safe, Secure, and Trustworthy Artificial Intelligence.”

This announcement will come on heels of one made this week by U.K. Prime Minister Rishi Sunak announcing the launch of an AI Safety Institute. Next month he'll be hosting an AI Safety Conference for global leaders in the U.K. (although not without controversy).

So perhaps not coincidentally, ahead of both these announcements, earlier this week OpenAI, Anthropic, Google, and Microsoft unveiled a new $10 million AI safety fund and the creation of the Frontier Model Forum, an organization designed to "advance safe AI development"

"The Frontier Model Forum…[will] share industry expertise with a wider audience as the Forum expands the conversation about responsible AI governance approaches."

Details are light right now, but expect more information (and more companies to join) as the focus of governments worldwide on AI safety intensifies.

More AI News This Week

• AWS announced that its 12% revenue growth this quarter is thanks in large part to generative AI.

• Quick read on how AirBnB is using AI to detect parties.

• Much longer (but excellent) read from the Financial Times on how generative AI will transform business.

• Nature is reporting about a new neural net that can learn language like a human, which could mean vast improvements over current LLMs like ChatGPT.

• If you want to learn about how Hello Fresh uses generative AI to create recipes at scale, go here.

• A new tool called Nightshade lets artists "poison" their work so that if a generative AI model picks it a tainted image to learn from it, the artist can, for example, make the model think a ball is cat, and thereby "infect" the model's data.

• It appears that Stability AI, creators of the popular AI generator model Stable Diffusion, is struggling.

Quote of the Week

"What was the PC about? It was about empowering you to be able to do things. Now, Copilot is in many ways like a PC, except that one can just go to this personal agent and say, Write me a document."

- Microsoft CEO Satya Nadella, to Business Insider

Time Well Spent: 5 AI Books Worth Reading

• The AI Revolution in Medicine: GPT-4 and Beyond 

• AI Made Simple: A Beginner’s Guide to Generative Intelligence

• The Coming Wave: Technology, Power, and the Twenty-first Century's Greatest Dilemma

• Is the Algorithm Plotting Against Us?: A Layperson's Guide to the Concepts, Math, and Pitfalls of AI

• AI 2041: Ten Visions for Our Future